[ejabberd] crash logs for RC2

Badlop badlop at gmail.com
Wed Nov 18 04:29:27 MSK 2009


2009/11/17 Jan Koum <jan.koum at gmail.com>:
>> You can provide one of the dumps to take a look.
>
> why just one?  i don't mind sharing them all:) i have scp'd them up to:
> http://camel.ethereal.net/~jkb/erl_crash_dumps/

Ohh, I can't download them:

Forbidden
You don't have permission to access
/~jkb/erl_crash_dumps/erl_crash_20091109-111819.dump on this server.


> but majority of our tcp
> connections do usually come from slow 3G mobile networks.

That may have some implication in the crashes.


> if it helps, i also see this stuff like this in the .log files and they all
> point at some SASL issue?  almost feels like some SASL timeout is causing
> badness?  or these "ERROR REPORT" safe to ignore just like various SSL
> connection handshake/timeout errors on the https server?

Some crashes take just 30 minutes to happen.
If those SASL errors are involved in the crash, there should be many
of them or be very harmful,
in order to take ejabberd down in 30 minutes.


> =ERROR REPORT==== 2009-11-16 07:53:30 ===
> ** State machine <0.5666.258> terminating
> ** Last event in was {xmlstreamelement,
>                          {xmlelement,"response",
>                              [{"xmlns","urn:ietf:params:xml:ns:xmpp-sasl"}],
>                              []}}
> ** When State == wait_for_sasl_response
> **      Data  == {state,
>                      {socket_state,gen_tcp,#Port<0.8797176>,<0.5664.258>},




---
Badlop
ProcessOne


More information about the ejabberd mailing list